Hi Michael,
thanks for your mail and sorry for the late reply. I've been away.
> I have been looking at ArgoUML (www.argouml.org) as a platform for writing a
> UI for an open-source O/R mapping tool
> (hibernate, OJB, Cayenne).
Daring, I'd say. I am not sure you'll be able to come up with a single,
model-based concept that enables all the plugins to generate the
framework-specific output; but then I've been wrong before. :-)
> 1. Integration with UML class diagrams - manipulation of diagrams
Good idea, although (IMHO) a bit overrated, since diagrams usually don't
scale unless you put in very advanced multi-aspect visualization
techniques. (I've been doing research on that in the past.. _hard_
problems). IMHO that is the #1 point why diagram-based tools don't cut
it for large scale development, which is really a pity.
> 2. Class model is XML (XMI format) (embedded in a zip archive)
XMI is definitely a good idea. Literally the 'last hope' for the tool
vendors to get their act together..
> 6. ArgoUML supports UML class diagram generation from source code
> and source code generation from UML models
Anything else would be useless, but you also need to make sure any plugins
take care of properly reengineering generated & modified source. I've
played with Poseidon (ArgoUML's commercial spinoff) and like it, but
haven't used it 'for real' to say whether it's worth the trouble.
(snip)
> Would this effort be welcome.? I realize that a great deal of effort has
> been expended on the cayenne modeler
> application, but if a redesign is being contemplated the above approach may
> offer a viable solution.
Yes and no - IMHO, of course. Modeler is a standalone solution and the
only one that Cayenne has right now; it would not make sense to just start
a rewrite for no good reason, just to end up with another dubious Swing
app. ;-)
Modeler is hard to modify but really not _that_ broken. My posting was a
step towards a more model-based mapping that would more easily enable
other frontends to be written; currently Modeler has to "know" that
"renaming this" or "setting that" requires you to also call a bunch of
other things as well; with a mostly self-maintaining mapping a lot of this
logic will hopefully simply go away, and this will pave the way for easier
Modeler maintenance and therefore more advanced features.
It will also pave the way for something that not only I would like to see,
namely an eclipse plugin. I confess: I'm an eclipse junkie since day one,
and one of the most interesting projects for Cayenne to hijack would be
the EMF, the eclipse modeling framework - and if you look at the pages
you'll see that they actually have XMI support, round-tripping, and a
lot more good things in the works. This is where I (emphasis on *I*!)
would like to take Cayenne/Modeler.
Still, to answer your question: would this effort be welcome? YES! By all
means, more exposure is always good, and an existing modeling solution is
of course a lot better than my EMF-based pipe dream.
Thanks for your mail!
Holger
This archive was generated by hypermail 2.0.0 : Fri Mar 21 2003 - 14:12:14 EST